The total accuracy of the Landsat-derived LULC data was 85.6, 89.6 and 90% with corresponding Kappa statistics of 82.7,
87.5 and 87.9% for MSS, TM and ETMþ, respectively, corroborating the standard accuracy of 85–90% for LULC mapping studies as recommended by Anderson et al. (1976). The application of rule-based post-classification refinement was found to be effective and improved accuracy by 10–12%. The MSS image had the lowest overall accuracy, which may be due to its coarse spatial resolution (Haack, 1987). Yang and Lo (2002) also noted that the problems associated with correctly classifying mixed pixels increases with decreasing image resolution, resulting in spectral confusion. In this study, spectral confusion was higher in the MSS image than in the TM/ETMþ images.